Although the number of emigrants in each education-experience cell is not directly observ-able, the available data allows me to construct sensible measures of emigration numbers for different skill groups. The idea behind the calculation is the following: take the gender-education-experience distribution from the Irish census and weight it with the corresponding numbers of workers who applied for PPS and NINo numbers in Ireland and the UK. By dividing the calculated emigrant number of a certain gender-education-experience cell by the number of people in Lithuania with the same characteristics, we obtain the emigration rates.
The calculation of emigration rates requires three assumptions about the emigrants’
gender-skill distribution: 1) the distribution is the same in the UK and in Ireland. 2) The distribution in 2002 is the same as in 2003, and 3) the distribution in 2005 is the same as in 2006.
The first assumption implicitly claims that no sorting behavior among mi-grants between the two destinations Ireland and the UK could be noticed. This assump-tion is backed by the recent literature on immigraassump-tion to Ireland and the UK. When we compare the descriptive statistics of the studies by Barrett & Duffy (2008, p.605) for Ireland and Dustmann et al. (2009, p.23) for the UK, the educational distribution of im-migrants from the A8 countries17 who came after 2004, looks fairly similar (see table 2).
Hazans & Philips (2009) analyze the occupational distribution of Lithuanians in Ireland and the UK. On the one hand, there is a difference in the sectors that employ Lithuanian immigrants in both countries. In the UK, around 30% of Lithuanian immigrants work
17 A8 countries are: Czech Republic, Estonia, Hungary, Latvia, Lithuania, Poland, Slovakia and Slove-nia.
in agriculture, whereas in Ireland this share is only 5%. This result could lead to the conclusion that migrants in the UK differed in their skills from those in Ireland. On the other hand, the same study shows that in both countries around 80% of Lithuanian migrants work in sectors that typically employ less-skilled workers, such as construction, health, trade, manufacturing, hotels and restaurants and agriculture. This indicates the absence of sorting behavior, so that it is reasonable to assume that the skill distribution of Lithuanian immigrants is the same in Ireland and the UK.
Assumptions 2) and 3) are reasonable as the education distribution among Lithuanian emigrants in Ireland did not change significantly from 2002 to 2006, even though the number of migrants is nine times higher in 2006. As we can see in table 1e), the share of immigrants with a third-level degree is slightly lower in 2006. At the same time, the share of those with lower secondary education is higher, but both distributions - 2002 and 2006 - do not differ a lot. Taken together, these three assumptions make it possible to extrapolate the skill distribution given in the Irish census to the UK and to the years that are not covered in the Irish census, 2003 and 2005. This allows me to present a more realistic picture of the size and impact of migration flows than we would get by only using the Irish data for 2002 and 2006 without extrapolating. In the robustness checks in section 5.2, I drop those assumptions. We will see that this has an impact on the magnitude, but not on the sign and statistical significance of the wage effects.
For the calculation of the number of emigrants for each gender-education-experience cell in the years 2002 and 2006, I use the number of Lithuanians in the Irish census of the same year and multiply it with a weighting factor, which accounts for the migration flows to the UK. For the years 2003 and 2005, I additionally weight the calculated number with the PPS and NINo numbers of those years.
Let xtghj denote the number of people in the Irish census of
gender(g)-education(h)-experience(j) cell at time t. For t = (2002, 2006), the calculated number of emigrants
where Mghjt is the calculated number of emigrants in cell ghj in year t. N IN Ot and P P St are the NINo and PPS numbers issued to Lithuanians in year t. The first term in parentheses (1 in this case), accounts for the fact that I consider the raw migrant numbers in the census 2002 and 2006 for Ireland. The second term in parantheses, N IN OP P S t
t , is a weighting factor for the extrapolation of the migrant skill distribution of the Irish census to the UK. If, for example, in 2006 the number NINo applications is twice the number of PPS applications, this factor is 2. Table 1e) displays the figures of PPS and NINo numbers issued between 2002 and 2006.
For the year 2003, I take the number of Lithuanian migrants in cell ghj of the year 2002 and weight it with the PPS and NINo numbers of 2003. This results in
Mghj2003 = x2002ghj P P S2003
P P S2002 weights the number of migrants in the Irish census in 2002 with the change in PPS numbers from 2002 to 2003. Suppose the number of Lithuanian immigrants in Ireland was 30% higher in 2003 than in 2002. Then P P SP P S2003
2002 = 1.3. N IN OP P S 2003
2002 accounts for the change in PPS numbers, as well as for the difference in migration flows to the UK and Ireland in 2003.18
The calculation of the number of emigrants in 2005 is analog the one of 2003:
Mghj2005 = x2006ghj P P S2005
2003 , which accounts for the size of migrant flows to the UK relative to Ireland and P P SP P S2003
2002, accounting for the change in migration flows to Ireland from 2002 to 2003. By multiplication of those two terms, P P S2003cancels out.
For my econometric analysis, emigration rates are more relevant than absolute emigrant numbers, as the coefficient δ in equation (1) can then be interpreted as a quasi-elasticity.
An increase in the emigration rate of one percentage point would then increase the real wage by δ percent.
The emigration rate mghjt for cell ghj in year t is
mghjt = Mghjt P
i
pghijt, (10)
where Mghjt denotes the number of emigrants calculated in equations (7) to (9). The denominator of equation 10 is the number of people in year t living in Lithuania and be-longing to cell ghj. Due to the fact that I do not have data covering the entire Lithuanian population, I have to calculate the number from the HBS. The HBS is representative at the household level, so that I can calculate the total number of Lithuanians in cell ghj by summing up the sampling weights pghijt19 over all observations i that are in cell ghj in year t.
19 The sampling weight pghijt is the inverse probability that observation i is included in the sample.
B Tables
Table 1: Descriptive Statistics
Year 2002 2003 2005 2006
a) Number of observations in the Lithuanian HBS, employees aged 18-64
All workers 3950 4136 4042 3874
Men 2322 2411 2426 2314
Women 1628 1725 1616 1560
b) Number of observations in the Irish census, employees aged 18-64
All workers 1904 - - 21779
Men 987 - - 12300
Women 917 - - 9479
c) Mean private sector income from employment in Litas, deflated by the HCPI. Source: own calculations from the Lithuanian HBS
All workers 1084 1142 1339 1533
Men 1139 1216 1405 1628
Women 906 905 1107 1249
d) Distribution of education in the Lithuanian HBS
lower secondary 9% 10.6% 10.9% 9.9%
upper secondary 68.8% 69.0% 67.5% 67.5%
third-level 22.2% 20.4% 21.6% 22.6%
e) Distribution of education of Lithuanians in the Irish census
lower secondary 16.7% - - 20.4%
upper secondary 63.4% - - 62.2%
third-level 19.9% - - 17.4%
f) Numbers of work permits (PPS and NINo).
Sources: Irish Department of Social and Family Affairs UK Department for Work and Pensions.
PPS 2709 2394 18680 16017
NINo 1430 3140 10710 24200
g) Lithuanian HCPI, 2005=100, source: Eurostat
97.334 96.291 100 103.788
Table 2: Distribution of education among A8 immigrants after 2004 in Ireland and the UK
authors Barrett & Duffy (2008) Dustmann et al. (2009)
country Ireland UK
lower secondary 11.1% 11.9%
upper secondary 61% 56.1%
third-level 28.2% 32%
Table 3: OLS, weighted with sampling weights. Men and women - private sector. De-pendent variable: log(real wage)
A: interaction region*year B: Controls FDI, Trade, GDP
(1) (2) (3) (4) (5) (6)
VARIABLES all interaction interaction all interaction interaction
male male*married male male*married
Emigration rate 0.595* 0.318 0.326 0.617** 0.447 0.345
[0.3071] [0.3549] [0.3843] [0.3024] [0.3120] [0.3299]
Emigration * Male 0.737** 1.078** 0.931*** 1.623***
[0.3431] [0.4169] [0.3420] [0.4881]
Emigration * married -0.379 0.116
[0.4614] [0.4621]
Emigration * married * male -1.013* -1.532**
[0.5812] [0.6380]
Male 0.165*** 0.145*** 0.142*** 0.164*** 0.142*** 0.137***
[0.0184] [0.0200] [0.0207] [0.0184] [0.0193] [0.0207]
Married 0.523*** 0.525*** 0.550*** 0.525*** 0.526*** 0.548***
[0.0251] [0.0250] [0.0292] [0.0249] [0.0248] [0.0293]
Children -0.036*** -0.036*** -0.033*** -0.036*** -0.035*** -0.032***
[0.0110] [0.0110] [0.0110] [0.0109] [0.0109] [0.0110]
Agglomeration 0.382*** 0.380*** 0.382*** 0.381*** 0.379*** 0.381***
[0.0232] [0.0232] [0.0231] [0.0228] [0.0229] [0.0227]
log(exports) 0.006 0.002 0.005
[0.0821] [0.0822] [0.0825]
log(gdp per cap.) 0.615* 0.613* 0.617*
[0.3164] [0.3171] [0.3177]
log(fdi inflows) 0.025 0.024 0.024
[0.0164] [0.0165] [0.0165]
Year Dummies yes yes yes yes yes yes
Education Dummies yes yes yes yes yes yes
Experience Group FE yes yes yes yes yes yes
Region Dummies no no no yes yes yes
Interaction yes yes yes no no no
Region*Year
Observations 9993 9993 9993 9993 9993 9993
Adjusted R2 0.3673 0.3677 0.3683 0.3667 0.3672 0.3679
Robust standard errors in brackets
*** p<0.01, ** p<0.05, * p<0.1
Table 4: Marginal effects of emigration on wages for different groups, results from table 3. P-values in brackets.
A: interaction region*year B: controls FDI, export, trade
All 0.5987* 0.6169**
(0.0525) (0.0426)
Women 0.3231 0.4473
(0.3641) (0.1532)
Men 1.058*** 1.3785***
(0.0012) (0.0002)
Women, unmarried 0.3303 0.3450
(0.3913) (0.2969)
Women, married -0.0448 0.4608
(0.6227) (0.4766)
Men, unmarried 1.4051*** 1.9768***
(0.0004) (0.0001)
Men, married 0.0198*** 0.5513***
(0.0031) (0.0005)
Table 5: Robustness checks. Marginal effects of emigration on wages for different groups.
P-values in brackets.
a) b) c) d) e)
All 0.2061 0.4617 0.5438* 0.4517* 0.8532***
(0.5688) (0.1358) (0.0823) (0.0652) (0.0090)
Women -0.0288 0.1931 0.2876 0.1973 0.6373*
(0.9273) (0.6143) (0.4217) (0.4119) (0.0754)
Men 0.8529*** 0.8788** 1.000*** 0.9896*** 1.4486***
(0.0050) (0.0121) (0.0020) (0.0065) (0.0003)
Women, unmarried 0.0136 0.1675 0.2885 0.2486 0.6629*
(0.9679) (0.6930) (0.4600) (0.3920) (0.0892)
Women, married -0.3538 -0.0177 -0.0690 -0.1340 0.5533
(0.6950) (0.9056) (0.6833) (0.6233) (0.1963)
Men, unmarried 1.2074*** 1.1510*** 1.3445*** 1.2495*** 2.1188***
(0.0018) (0.0039) (0.0005) (0.0079) (0.0001)
Men, married -0.1034*** 0.0255** -0.0463*** 0.1499** 0.6376***
(0.0056) (0.0158) (0.0031) (0.04251) (0.0004)
a) Emigration rates of other cells included (section 5.2.3) b) experience included as a continuous variable (section 5.2.4) c) interaction education group * year (section 5.2.5)
d) 2-year experience cells (section 5.2.4) e) 10-year experience cells (section 5.2.4)
C Figures
Figure 1: Emigrant shares after EU accession: number of work permits in the UK and Ireland from 2004-2007 divided by the number of employed people in the source country in 2003. Source: Eurostat.
Figure 2: Number of Lithuanian emigrants to the UK and Ireland, measured by registra-tion for work permits, i.e. PPS and NINo numbers, 2002-2007. Sources: Irish Department of Social and Family Affairs, UK Department for Work and Pensions
10000 15000 20000 25000 30000 35000
PPS NINo
0 5000 10000 15000 20000 25000 30000 35000
2002 2003 2004 2005 2006 2007
PPS NINo
Figure 3: Lithuania: real GDP per capita, real average wages, unemployment. Source:
2002 2003 2004 2005 2006 2007
Real GDP Unemployment Real Wages
Figure 4: Scatter: wages and emigration rates for different groups (male and female, married and unmarried. Source: own calculations.)
Figure 5: Wage increases for different groups, 2002-2006, 2005=100. Source: own calcu-lations, based on the Lithuanian HBS.
Figure 6: Gross fixed capital formation in million Litas. Sources: IMF International Financial Statistics
10000 15000 20000 25000 30000
0 5000 10000 15000 20000 25000 30000
1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008
Institute for International Integration Studies